Article pubs.acs.org/JPCC
Mechanisms of Photogeneration and Relaxation of Excitons and Mobile Carriers in Anatase TiO2 Maria C. Fravventura, Laurens D. A. Siebbeles, and Tom J. Savenije* Opto-electronic Materials Section, Department of Chemical Engineering, Delft University of Technology, Julianalaan 136, 2628 BL, Delft, The Netherlands S Supporting Information *
ABSTRACT: We studied the temperature dependence of photophysical products on UV excitation of smooth dense anatase TiO2 with frequency- and time-resolved microwave conductance measurements. At 100 K, we observed the subnanosecond formation of microsecond-lived self-trapped excitons (STEs) with a 65% quantum yield, irrespective of the UV excitation wavelength. The remaining 35% photoexcitations results in mobile conduction band electrons (e−CB); the yield of e−CB gradually increases up to about 100% at 300 K. The complex mobility of e−CB is independent of temperature between 100 and 300 K. We explain the temperature-dependent quantum yield of e−CB by the thermally activated escape of the electron from its positive counter charge. At low temperature, the Coulomb attraction causes the electron to remain at short distance from the hole forming a neutral pair. This pair stabilizes by inducing a relaxation of the surrounding lattice resulting in a self-trapped exciton. The excess polarizability volume of each STE is found to be temperature independent and of the order of 104 Å3. Our results indicate that not only in dense smooth TiO2 but also in nanostructured TiO2 STEs are efficiently photogenerated at low temperature. For the first time we provide quantitative information about the quantum yield of excitons in bulk anatase TiO2, the existence of which has been previously demonstrated only qualitatively by means of photoluminescence measurements.
1. INTRODUCTION
2. EXPERIMENTAL SECTION 2.1. Sample Preparation. Thin films of Pol-TiO2 of 90 ± 10 nm thickness were purchased from Everest Coatings (Delft, The Netherlands). The films were prepared by chemical vapor deposition on top of 1 mm thick quartz plates. Subsequent annealing at 450 °C for 2 h was performed to improve stoichiometry. The crystallite size is 90 ± 10 nm as determined from atomic force microscopy (AFM) and X-ray diffraction (XRD) measurements (see Figure 1S in Supporting Information). A dye-sensitized Pol-TiO2 sample (DS-Pol-TiO2) was prepared by spin-casting a solution of the polymer MDMOPPV (poly[2-methoxy-5-(3′,7′-dimethyloctyloxy)-1,4-phenylene-vinylene] (Sigma Aldrich) dissolved in chlorobenzene on top of a bare Pol-TiO2 film. This sample allows us to determine the quantum yield for exciton and charge carrier generation on excitation with visible light. In addition, a 2 μm thick nanocrystalline TiO2 (NC-TiO2) film was prepared by doctor blading a paste of 9 nm anatase TiO2 nanoparticles on a quartz substrate according to a procedure presented elsewhere.9 This latter sample contains a high density of electron traps (ca. 1018 cm−3)10 and it can be used to determine the individual contribution of holes to the photoconductance. 2.2. Microwave Photoconductance Technique. The FTRMC technique and analysis method used in this work have
Anatase titanium dioxide (TiO2) is a wide band gap semiconductor used for a variety of purposes including photocatalysis and solar energy conversion.1−4 In order to optimize the performance of TiO2-based opto-electronic devices, detailed understanding of the mechanisms of charge generation and transport is essential. At room temperature, photoluminescence from anatase TiO2 is extremely weak, suggesting that thermal energy (25 meV) is sufficient to induce dissociation of the excited state into separate carriers. In contrast, at low temperature enhanced sub-bandgap photoluminescence is observed.5−8 Hence, at low temperature part of the photoexcitations leads to long-lived bound electron−hole pairs (excitons). In the present work, we investigate the temperature dependence of the populations of excitons and mobile charge carriers generated in polycrystalline anatase TiO2 (Pol-TiO2) on UV photoexcitation. We use the frequency- and time-resolved microwave conductance (FTRMC) technique and vary the temperature of the sample from 100 to 373 K. Previously, we found that UV excitation of a variety of TiO2 samples at 300 K leads to free and trapped charge carriers.9 Here, we show that in addition to charge carriers excitons are also formed with a quantum yield that increases to 65% as the temperature is reduced to 100 K. To our knowledge, no quantitative information regarding the yield of excitons in TiO2 has been reported previously. © 2014 American Chemical Society
Received: January 6, 2014 Revised: March 18, 2014 Published: March 19, 2014 7337
dx.doi.org/10.1021/jp500132w | J. Phys. Chem. C 2014, 118, 7337−7343
The Journal of Physical Chemistry C
Article
Figure 1. Frequency- and time-dependent experimental (left) and modeled (right) ΔP(ω,t)/P contour plots for 300 nm excitation of Pol-TiO2 at an absorbed photon fluence of IA = 1.4 × 1011 photons/cm2 at 300 K (a,d), 200 K (b,e), and 100 K (c,f).
The global evaluation of the ΔP(ω,t)/P signal in the −30 and +30 MHz frequency offset range allows the investigation of the microwave photoconductance with 1 ns time resolution owing to the off-resonance fast response of the system. Quantitative information about the quantum yield and complex mobility of photophysical products can be obtained by theoretical modeling of the experimental frequency- and time-dependent ΔP(ω,t)/P contour plots as the electrical response of an RLC circuit.9,13 The time-dependent concentrations of photophysical products are considered as input parameters and are adjusted to obtain an optimal fit of the modeled ΔP(ω,t)/P to the experimental data. Each photophysical product has a characteristic ratio of its contribution to the imaginary and real components of the photoconductance, γ = |ΔG″/ΔG′|. Usually γ < 1 for mobile charge carriers, while γ > 1 is typical for polarizable products such as bound electron− hole pairs, localized trapped charges or charges confined in a bound domain. In the case of a purely real photoconductance (ΔG = ΔG′), the ΔP(ω,t)/P contour plot is symmetric around ωRES, while in the case of a purely imaginary photoconductance (ΔG = jΔG″) it is fully antisymmetric. In general, the ΔP(ω,t)/ P contour plot exhibits a superposition of symmetric and antisymmetric features9,13 due to a complex photoconductance with real and imaginary components that are both nonzero.
been fully described previously.9 The methodology allows us to determine the microwave complex photoconductance, ΔG = ΔG′ + jΔG″, with j2 = −1.10−12 Briefly, the sample of interest is mounted in a resonant X-band cavity and photoexcited by a nanosecond laser pulse (3 ns full width half-maximum and 10 Hz repetition rate, which is sufficiently low for all charges to recombine between successive pulses). Formation and decay of photophysical products such as mobile charge carriers and excitons are probed by monitoring the change in microwave power using a sensitive microwave detection setup. The normalized change in microwave power, ΔP(ω,t)/P is proportional to the complex photoconductance ΔG(ω,t), according to ΔP(ω , t ) = −K (ω)ΔG(ω , t ) P
(1)
In eq 1, K(ω) is the frequency-dependent sensitivity factor of the system, which depends on the geometry and quality factor of the resonant cavity containing the sample. By measuring microwave transients at different frequencies between −30 and +30 MHz from the resonance frequency of the loaded cavity, a ΔP(ω,t)/P contour plot is obtained. The horizontal black line at zero frequency-offset in the ΔP(ω,t)/P plot (see Figure 1) denotes the resonance frequency (ωRES) of the loaded cavity. At this frequency, the formation of a standing wave within the cavity results in multiple interrogations of the sample by the microwave probe. In this condition, we thus achieve the highest sensitivity for the real component of photoconductance (ΔG′), however, at the cost of an increased response time of 18 ns. Away from ωRES, the relaxation of the resonance condition results in progressively fewer interactions of the microwaves with the sample, resulting in a reduction of the system response time to about 1 ns (see Figure 2S in Supporting Information).
3. RESULTS AND DISCUSSION Figure 1 shows the frequency- and time-dependent ΔP(ω,t)/P contour plots recorded upon 300 nm excitation of Pol-TiO2 at different temperatures. At 300 K (Figure 1a), the plot is highly symmetric around ωRES, and all ΔP(ω,t)/P values are positive, which is in agreement with previously published data on PolTiO2.9 On comparing Figure 1a−c, it can be observed that lowering the temperature induces a strong asymmetry around 7338
dx.doi.org/10.1021/jp500132w | J. Phys. Chem. C 2014, 118, 7337−7343
The Journal of Physical Chemistry C
Article
ωRES. The maximum ΔP(ω,t)/P decreases and its position shifts from ωRES at 300 K to lower frequencies (−5 MHz) at 200 and 100 K. In addition, Figure 1b,c shows negative ΔP(ω,t)/P values in the positive frequency-offset regime. To deduce the yield and rate constants for formation and decay of photophysical products, the contour plots were modeled as described in the Experimental Section. In agreement with previous findings,9 only one photophysical product with γ = 0.16 is required to model the ΔP(ω,t)/P plot of Figure 1a. In analogy with previous reasoning,13 we attribute this microwave response to formation of conduction band electrons (e−CB) with a quantum yield close to 100%.
Figure 3. Temperature dependence of the quantum yield of conduction band electrons, φe−CB upon 300 nm photoexcitation. The inset shows the temperature dependence of the quantum yield, ϕX upon 300 nm excitation (filled squares against the left axis) and of the luminescence intensity reported by Tang et al.5 for 360 nm excitation of pure anatase (empty squares against the right axis).
subnanosecond time scale and decays with a monoexponential temperature-independent lifetime τc = 3.3 ± 0.2 μs. From the transients given in Figure 2, we can observe that the decay of X does not lead to an increase of e−CB. We thus propose that at low temperatures formation of X occurs at the expense of e−CB. On the basis of the experimental results discussed so far, we cannot conclude that e−CB and X are the only photophysical products in the material. Other photophysical products that do not give appreciable contribution to the complex photoconductance might also be generated and be silent in the FTRMC experiment. This aspect will be discussed later in more detail. However, at all temperatures an upper limit for the quantum yield of photogeneration of X is given by φX = (1 − φe−CB) and it is displayed in the inset of Figure 3 as function of temperature. Table 1 lists the quantum yields and the kinetic parameters used in the theoretical RLC model to reproduce the experimental ΔP(ω,t)/P data. Furthermore, we also measured the temperature dependent ΔP(ω,t)/P signal upon 250 and 340 nm excitation (see Figure 3S in Supporting Information), obtaining results comparable to those for 300 nm excitation. This indicates that irrespective of the excitation wavelength the quantum yields of e−CB and X vary with temperature as displayed in Figure 3 and listed in Table 1. As mentioned above, X might a priori be attributed to a variety of photophysical products such as bound electron−hole pairs, charge carriers confined in a bound domain, or deeply trapped in the lattice. In order to clarify the nature of X, we carried out two additional experiments. (I) We photoexcited a dye-sensitized TiO2 film (DS-Pol-TiO2) at 500 nm, corresponding to the absorption peak of the dye. Note that the TiO2 film used for the sensitized sample is equivalent to the Pol-TiO2 sample already discussed and consists of a dense, smooth anatase layer. Upon 500 nm excitation electrons are injected from the LUMO level of the dye into the conduction band of TiO2 on subnanosecond time scale.15,16 Hence, in contrast to UV excitation of the bare Pol-TiO2 film, 500 nm excitation of DS-Pol-TiO2 results in photogenerated holes residing in the HOMO level of the dye, rather than in the valence band of TiO2.17 Interestingly, the experimental ΔP(ω,t)/P contour plots of DS-Pol-TiO 2 (see Figure 4S in Supporting Information) do not exhibit any temperature dependence in the extent of the asymmetry around the resonance frequency.
Figure 2. Modeled time-dependence of the densities of e−CB produced in Pol-TiO2 upon 300 nm excitation at different temperatures. The inset shows the time dependence of the normalized densities of the photophysical product X for T ≤ 250 K.
Figure 2 shows the time-dependent densities of conduction band electrons, Ne−CB, used to obtain the theoretical frequencyand time-dependence of ΔP(ω,t)/P in Figure 1d−f.14 Interestingly, both the formation time and the maximum of Ne−CB are temperature dependent. Whereas at 373, 300, and 250 K, e−CB is formed on subnanosecond time scale, at lower temperatures the generation of e−CB is progressively delayed with characteristic formation times of 2 ± 1, 5 ± 1, and 8 ± 1 ns for 200, 150, and 100 K, respectively. The decay kinetics do not significantly vary with temperature and can be represented by a double-exponential profile with lifetimes τa = 7 ± 1 μs and τb = 45 ± 5 μs. The corresponding quantum yields for formation of mobile conduction band electrons are derived from the maxima of the Ne−CB transients in Figure 2 and from the absorbed photon fluence. Figure 3 shows the temperature dependence of φe−CB, which deviates from the exponential function expected for a single activation energy. However, the limited absolute accuracy − in the determination of φeCB (±5%) resulting from our experimental and modeling approach disables a full quantitative analysis. Most interestingly, the ΔP(ω,t)/P contour plots recorded at temperatures lower than 300 K (Figure 1b,c) could not be modeled using only e−CB. The addition of a second photophysical product (denoted as X) characterized by a large imaginary photoconductance (γX = 30.8) was required. The presence of X accounts for the strongly antisymmetric features in Figure 1b,c. The normalized time-dependent densities NX used for the modeling are shown in the inset in Figure 2 for T ≤ 250 K. In this temperature range, X is formed on a 7339
dx.doi.org/10.1021/jp500132w | J. Phys. Chem. C 2014, 118, 7337−7343
The Journal of Physical Chemistry C
Article
Table 1. Quantum Yields (±5%), Formation- and Decay-Times of e−CB and X Used in the RLC Model to Reproduce the Experimental Contour Plots in Figure 1 373 K
300 K
250 K
200 K
150 K
100 K
φe−CB
>95%
>95%
85%
68%
40%
35%
formation e−CB (ns) decay e−CB (μs)